| Commit message (Collapse) | Author | Age |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
The filepath changes with every flag change. It is thus easier to only
store a limited path that remains stable, and figure out the rest as
the property is requested (we'll have to translate it anyways once we
the file handoff protocol is implemented).
The reason why we don't update the mimeMessage path on every
modification is because we move the message during change replay,
and not while storing the modification in the db. This would lead to
the message-path on disk not correspond to what is in the db for some
time.
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
(except for documentation).
|
| | |
|
| |
|
|
|
| |
I'm not sure we require this cache at all, and we arent using it
properly in the current resource.
|
| |
|
|
|
| |
The sync and changereplay can not run at the same time, or would have
to share the transaction otherwise.
|
| | |
|
| |
|
|
|
| |
The subfolder implementation is not according to any standard it seems,
but at least it works for now.
|
|
|
Respectively a first prototype thereof.
|
